> ## Documentation Index
> Fetch the complete documentation index at: https://docs.axottle.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Egress Guard

> Control of dangerous outbound ports on nodes: fixed catalog of categories, application status and drift reconciliation.

Egress Guard is a deliberately compact control of dangerous outbound traffic from the node. You turn on and off categories from a fixed catalog of dangerous ports; the panel compiles the selection into a versioned rule set with a control hash, the agent sets the rules on the node and returns the hash of what **is actually** installed in the kernel. The panel considers the protection applied only when the desired and actual hashes matched.

> [!NOTE]
>
> For whom: Panel Administrator / SRE. Required rights: Administrator session; node requires an agent with firewall support.

> [!NOTE]
>
> Don’t confuse Egress Guard with “external transit egress blocks” – this is another feature: VLESS hops in the Flow column, see [Routing Flow](https://docs.axottle.com/en/connection-units/routing.md).

## Principle "Save  ⁇  applied"

The changes first save the desired policy, and the status of its application by node is displayed separately in the **Apply** tab.

## Category catalogue

The directory is fixed - there is no free list of ports in the standard interface, and it is consciously:

| Category | Ports | Default Action |
| - | - | - |
| Mail SMTP | TCP 25, 465, 587, 2525 | Locked (reject) - **enabled by default** |
| UDP amplification | UDP 17, 19, 69, 111, 137, 161, 389, 1900, 11211 | Locked (drop) – enabled by default |
| Remote Administration (SSH/RDP/VNC) | - | Detection only |
| Databases | - | Detection only |
| IoT Scanning Purposes | – | Detection Only |
| Proxyware / Mining / Suspicious | - | Detection only |

* Actions: `drop`, `reject`, `detect` (detection does not put a rule in the firewall - only powers the detectors).
* **DNS (53) and NTP (123) are never consciously blocked by**, instead of being blocked by an anomalous fan-out detector.
* Port redefinition is only available in extended mode; policies with unknown categories, activities, or ports are rejected.

## How reconciliation works

* The policy is stored with a single entry with the version (each save raises the version) and the `auto_reapply` flag.
* About every 5 minutes, the panel bypasses all the nodes: if the desired hash has changed or the rules have not yet been applied, it sends an application; otherwise it performs a cheap state reading that catches the drift that occurred outside the panel (manual rule reset, node reboot).
* **Auto repair drift is performed only when switched on `auto_reapply`** - and never in the same aisle that he just applied the rules (cycle protection), when the flag is turned off, the drift is simply fixed and left to the operator's decision.
* A node whose agent does not support firewall operations is shown as `unsupported`, not "applied."
* Node statuses: `pending`, `applied`, `drift`, `failed`, `unsupported`, `unknown` are sorted from worst to best in the interface. Policy changes and reconciliation actions are written in the audit log.

## Setting up

**1. See the directory and current policy**

Open **Security by Egress Guard**.

**2. Keep the policy.**

Enable the desired categories and click **Save policy**.

**3. Apply now (if desired)**

Go to the tab **Application** and click **Convert now**.

## Verification

* After changing the policy, wait for the scheduled reconciliation or click **Contact nowZ**.

> [!TIP]
>
> The protection is valid on the node if its status - `applied`That is, the hash of the rules in the nucleus of the node coincided with the desired.

## Rollback

* Unlock the lock: Turn off the category or transfer it to `detect`, save the policy and click **Contact now**.
* Auto repair drift can be turned off with the `auto_reapply` flag - the panel will cease to arbitrarily return the rules and will only report drift.

## Limitations and risks

* `detect` categories don’t protect traffic – they only observe.
* There is a window (up to one reconciliation cycle) between maintaining the policy and applying it.
* On nodes without firewall support, the agent does not have protection - the status of `unsupported` requires updating or reconfiguring the agent.

> [accordion] Node hangs in drift
>
> Open the tab **Application** and click **Confirm Now**. If auto repair is enabled, the panel will reapply the desired state.

> [accordion] Node shows off unsupported
>
> The agent on the node has no firewall capability. Check the agent's version and capabilities on the node page.

> [accordion] Maintained policy, and status is still pending
>
> This is normal immediately after saving. Wait for the reconciliation cycle or press **Contact now**.

Related pages: [Routing Flow](https://docs.axottle.com/en/connection-units/routing.md), [Journal and audit](https://docs.axottle.com/en/panel/journal-and-audit.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.